Add under-800M Danish ASR benchmark comparison

README.md

@@ -124,6 +124,27 @@ The NeMo checkpoint processed the benchmark at **795.8x realtime** on one A100.
 The official run saved raw per-sample outputs, and an independent offline
 rescore reproduced 12.37% exactly.
 
+### Comparison with models below 800M parameters
+
+![Mean WER of Ekko v1 Tiny and all twelve open-weight leaderboard models below 800M parameters](comparison.png)
+
+The chart combines Tiny's Q5_0 release score with all 12 open-weight models
+with reported parameter counts below 800M and results for all five test sets in the
+[leaderboard snapshot of 4 October 2026](https://huggingface.co/datasets/RyeAI/danish-asr-leaderboard/tree/d5445121bb307df26ccf58d2a1fc04ae387ec38a/results).
+Tiny is not yet a submitted row in that snapshot. The scores are equal-weight
+means across five benchmark test sets totaling 26,780 rows. Tiny's 12.37% is 0.27
+percentage points below the next entry, Svale 600M at 12.64%. These are point
+estimates from this snapshot; the leaderboard may change as models are added.
+Parameter labels use the sizes advertised on the leaderboard where available;
+Tiny uses its published 110M size.
+
+<details>
+<summary>Accuracy against parameter count</summary>
+
+![Mean WER versus model parameter count for the same thirteen models](size-vs-accuracy.png)
+
+</details>
+
 ### Native GGUF quantization
 
 The F32, F16, Q8_0, Q6_K, Q5_K, Q5_0, Q4_K, and Q4_0 files are